4 ms·
> "The AI is 'punished' for wrong answers ('I love racism') and 'rewarded' for right answers ('As a large language model trained by OpenAI, I don’t have the abi
by smnrg 4y ago
> "The AI is 'punished' for wrong answers ('I love racism') and 'rewarded' for right answers ('As a large language model trained by OpenAI, I don’t have the ability to love racism.')"
Sounds like we train AIs like German Shepherds.